Robust Coarse-to-Fine Registration Scheme for Mobile Laser Scanner Point Clouds Using Multiscale Eigenvalue...

Area of Science:

• Robotics
• Computer Vision
• Geomatics

Background:

• Pairwise registration of mobile laser scanner (MLS) point clouds faces challenges with corresponding point identification and registration matrix accuracy.
• Existing methods struggle with aligning diverse MLS data frames effectively.

Purpose of the Study:

• To develop a robust coarse-to-fine registration method for aligning different frames of MLS point clouds into a unified coordinate system.
• To enhance the accuracy and reliability of MLS point cloud registration.

Main Methods:

• A multiscale eigenvalue statistic-based descriptor was optimized for feature description.
• A weighted covariance matrix was constructed using local geometric point distribution for feature extraction.
• Corresponding point pairs were identified in feature space, with outliers removed using geometric consistency constraints.
• Coarse registration was performed using identified point pairs, followed by fine registration using the Iterative Closest Point (ICP) algorithm with the coarse result as initialization.

Main Results:

• The proposed method successfully aligned MLS point clouds from different frames with high accuracy.
• Experimental results on Autonomous Systems Lab (ASL) Datasets demonstrated superior performance compared to existing registration methods.
• The optimized descriptor and geometric consistency constraints effectively improved registration robustness.

Conclusions:

• The developed coarse-to-fine registration method offers a significant improvement for MLS point cloud alignment.
• This approach provides a more accurate and reliable solution for integrating data from multiple MLS scans.
• The findings contribute to advancements in 3D data processing for autonomous systems and geospatial applications.